Full-length cDNA clones for human DDX3X (BC011819) and DDX3Y (BC034942) were purchased from GE-Dharmacon and subcloned into a pCS2+ expression vector with a C-terminal HA or FLAG tag. Primers used in the subcloning are listed in Supplementary Table S1. Pharmacological inhibition of transcription was performed with actinomycin D (Sigma-Aldrich A1410). Primary antibodies used in this study include mouse anti-DDX3X (Santa Cruz sc-81247; 1:1,000), rabbit anti-DDX3X (Bethyl Laboratories A300-474A; 1:1,000), rabbit anti-DDX3X (Aviva OAAB01241; 1:1,000), rabbit anti-HA (Cell signaling Technology C29F4; 1:2,000), rabbit anti-GFP (Invitrogen A-11122,1:1,000) and custom-made rabbit DDX3Y antibody (1:2,000) (Gong et al., 2021). Secondary antibodies include HRP-conjugated mouse anti-β-actin (Cell Signaling Technology 8H10D10, 1:5,000), HRP-linked horse anti-mouse IgG (Cell Signaling Technology 7076, 1:10,000) and HRP-linked goat anti-rabbit IgG (Cell Signaling Technology 7074, 1:10,000).

U87MG cells (ATCC) were cultured in MEM (Corning) supplemented with 10% fetal bovine serum (FBS; Gibco) at 37 °C with 5% CO₂. HCT116 cells (ATCC) were cultured in McCoy’s 5A (ATCC) supplemented with 10% fetal bovine serum (FBS, Gibco) at 37 °C with 5% CO₂. Cells were transfected with 10 nM DDX3X, DDX3Y or control siRNA (Dharmacon J-006874-06-0002, L-011904-01-0005, and D-001810-01-05) at 50% confluency using Lipofectamine RNAiMAX (Invitrogen), or with plasmids at 70% confluency, using Lipofectamine 3000 (Invitrogen). For experiments that used both siRNA and plasmid, siRNA was transfected first for 24 h before plasmids being transfected for another 48 h.

Small guide RNAs (sgRNAs) were designed to target exon 5 or exon 1 of the DDX3X gene using CRISPRscan. Protospacers DDX3X-g1 and DDX3X-g2 were cloned into a pSpCas9(BB)-2A-GFP/PX458 expression vector (Addgene 48138) using a scarless golden gate cloning strategy as described (Ran et al., 2013). Primers used in subcloning the protospacers are listed in Supplementary Table S1. Plasmid was transfected in U87MG cells at 500 ng/mL for 48 h followed by fluorescence-activated cell sorting (FACS) to sort GFP-expressing cells and plated individually into 96 well plates. Clones were cultured in MEM supplemented with 20% FBS and expanded for 2-3 months to collect enough samples for subsequent procedures. Genomic DNA was isolated and then PCR amplified using genotyping (GT) primers followed by PCR cloning (NEB E1202S) and Sanger sequencing. Total RNA was isolated using the RNeasy Mini Kit (Qiagen) and then reverse transcribed into cDNA using the iScript cDNA Synthesis Kit (Bio-Rad Laboratories) with DNase I (Qiagen) treatment. With the cDNA as the template, PCR amplification was performed, followed again by PCR cloning and Sanger sequencing.

Cells were lysed on ice in RIPA lysis buffer (Fisher Scientific 89901) with Halt protease inhibitor cocktail (ThermoFisher 78429) and Halt phosphatase inhibitor cocktail (ThermoFisher 78420) added, and processed for western blotting as described (Li et al., 2018). Immunoprecipitation (IP) was carried out with Pierce™ Anti-HA Agarose (ThermoFisher 26181) at 4 °C overnight with the bound protein released by boiling in Laemmli SDS sample buffer (ThermoFisher J61337-AD) at 95 °C for 10 min. Blots were detected with HRP-conjugated antibodies and Clarity Western ECL substrate (Bio-Rad #1705061) using a Bio-Rad ChemiDoc touch imager. Total RNA was extracted from collected embryos using RNeasy Mini Kit (Qiagen 74104) and then reverse-transcribed into cDNA using iScript cDNA Synthesis Kit (Bio-Rad 170-8891). Using qMax Green qPCR Mix (Accuris ACC-PR2000-L-100), quantitative PCR (qPCR) was carried out on Quant Studio 6 Flex Real-Time PCR system (Applied Biosystems). The cycling conditions include an initial denaturation step at 95 °C (2 min), 40 cycles of 5 s at 95 °C and 30 s at 60 °C. Primers used for RT-qPCR are listed in Supplementary Table S1.

Previous studies have shown the upregulation of Ddx3y mRNA upon knockout of Ddx3x in mice (Hoye et al., 2022; Patmore et al., 2020). However, it remained unclear how the translated Ddx3y protein is affected, due to the lack of an antibody that can differentiate between Ddx3x and 3y. With a recently generated antibody specifically recognizing human DDX3Y but not 3X (Gong et al., 2021), we were able to use cultured human cell lines to examine the regulation of DDX3Y protein by DDX3X. In HCT116 colon cancer cells, we detected 44% increase in endogenous DDX3Y following siRNA-mediated knockdown of DDX3X (Figure 1A). This was accompanied by 98% increase in DDX3Y mRNA (Figure 1B), suggesting that the upregulation of DDX3Y protein was primarily caused by elevated transcripts. The smaller changes in protein as compared to mRNA may reflect additional negative feedback at translational or post-translational level. As DDX3X has been shown to function in various processes of RNA metabolism, including transcription, nuclear export and stability (Mo et al., 2021; Soto-Rifo and Ohlmann, 2013), we assessed if the stability of DDX3Y mRNA can be altered by DDX3X. After treating the cells with the transcription inhibitor actinomycin D, we measured the amount of DDX3Y mRNA over time (Figure 1C). Upon knockdown of DDX3X, DDX3Y mRNA was degraded more slowly with a half-life of 4.3 h as compared to control, which had a half-life of 3.0 h (Figure 1D). In addition, after 8-h treatment with actinomycin D, there was a significant difference in the fold change of DDX3Y mRNA between control and DDX3X knockdown cells (Figure 1E), implying that DDX3X destabilizes DDX3Y mRNA. However, we cannot rule out the possibility that DDX3X also represses DDX3Y transcription.

We next turned to U87MG, a glioma cell line. To knock out DDX3X in these cells, we used two guide RNAs (gRNAs) individually for CRISPR/Cas9-mediated genome editing and obtained one clone with each gRNA (Supplementary Figure S1). Because the cDNA sequences of DDX3X and 3Y are highly homologous to each other, both gRNAs were designed to contain mismatches for DDX3Y so that this paralog was unlikely to be targeted (Supplementary Figure S2). Sequencing of both the genomic DNA and the cDNA indicated that Clone 1 (obtained with gRNA-1) contained a 159-bp in-frame deletion of the entire Exon 5 in the coding sequence of the mRNA, which resulted in a smaller DDX3X protein with residues 96-148 deleted (Supplementary Figure S3). In contrast, Clone 2 (obtained with gRNA-2) had a 14-bp deletion that was confirmed by sequencing of the DDX3X cDNA (Figure 2A). Although our initial western blotting with a Bethyl anti-DDX3X antibody (raised against residues 1-50) suggested that DDX3X was completely knocked out in Clone 2, another antibody from Santa Cruz (raised against the broader N-terminus region) picked up a band with a size similar to that of DDX3X but a significantly lower intensity (∼34% of parental cells; Figure 2B). We originally suspected that this band represented DDX3Y, as the two paralogs share high sequence homology. However, western blotting using the DDX3Y-specific antibody failed to obtain any signal, and RT-qPCR amplification curve for DDX3Y in either the parental cells or Clone 2 did not plateau within 40 cycles (data not shown), suggesting that DDX3Y expression may have been permanently silenced in these cells. In addition, this band was reduced when the cells were transfected with an siRNA targeting DDX3X but not another one targeting DDX3Y (Figure 2C), indicating that it represented DDX3X. A careful re-examination of the DDX3X cDNA revealed that the cells likely utilized a slightly downstream and originally out-of-frame AUG to restore the open reading frame after the deletion (Figure 2A), thereby retaining most of DDX3X protein sequence and probably function. This suggests the existence of selection pressure against loss of DDX3X function. Indeed, a search against the DepMap portal (Tsherniak et al., 2017) did infer a negative impact of DDX3X loss on U87MG cells (Supplementary Figure S4). The lower endogenous DDX3X protein levels in Clone 2 are probably due to translational repression by the upstream AUG, the original start codon that became out of frame in this clone (Figure 2A). A similar, naturally occurring translational repression by upstream AUGs has been reported for DDX3Y (Jaroszynski et al., 2011).

Since the endogenous DDX3Y was not detectable in either Clone 2 or the parental U87MG cells with the specific antibody used in Figure 1A (not shown), we transfected these cells with a plasmid encoding C-terminally HA-tagged DDX3Y. Western blotting with an anti-HA antibody showed higher levels of the exogenous DDX3Y protein with a fold change of ∼6.9 in Clone 2, as compared to parental cells transfected with the same amount of plasmid (Figure 3A). To test if this effect was caused by reduced DDX3X levels in Clone 2 (Figure 2B), we knocked down DDX3X in the parental U87MG cells and also observed a significant increase in the exogenous DDX3Y protein (Figure 3B). RT-qPCR result of mRNA extracted from the same batch of cells indicated that cells expressing exogenous DDX3Y contained 300-fold more DDX3Y mRNA than mock-transfected cells (data not shown). Thus, the total DDX3Y mRNA essentially represented the exogenous transcripts. No significant change in DDX3Y mRNA was found upon DDX3X knockdown (Figure 3C), suggesting that DDX3X regulates DDX3Y at protein level, such as translation or turnover. To examine the potential effect of DDX3X on DDX3Y turnover, we treated cells with the translation blocker cycloheximide (CHX) and monitored the degradation of exogenous DDX3Y. In parental U87MG, DDX3Y protein became undetectable within 4 h of CHX treatment. By contrast, in Clone 2 cells, where there was reduced endogenous DDX3X, degradation of DDX3Y protein was greatly slowed down (Figure 3D). Interestingly, DDX3X was much more stable despite the high sequence homology to DDX3Y, with over 80% of protein remained even after 12 h of CHX treatment (Figure 3D). These results indicate that DDX3X facilitates the turnover of DDX3Y protein in U87MG cells.

Both DDX3X and the yeast homolog Ded1p have been shown to form oligomers to unwind RNA duplexes cooperatively (Putnam et al., 2015; Song and Ji, 2019). Based on the high sequence homology between DDX3X and 3Y, we hypothesized that these two paralogs interact with each other. To test this hypothesis, we co-expressed C-terminally HA-tagged DDX3Y and FLAG-tagged DDX3X in Clone 2 cells and performed co-IP using agarose beads conjugated with an anti-HA antibody. As shown in Figure 4A, Flag-tagged DDX3X was pulled down along with the HA-tagged DDX3Y. Likewise, endogenous DDX3X also co-precipitated with HA-tagged DDX3Y in parental U87MG cells (Figure 4B), further confirming the binding of DDX3X to DDX3Y.

Post-translational modifications on lysine residues, such as ubiquitination, are well known to trigger protein degradation (Damgaard, 2021). Among the 32 lysine (K) residues on DDX3X, three (K55, K138, K162) have been linked to ubiquitin-mediated degradation (Wang et al., 2021); however, all three are shared with DDX3Y and are therefore unlikely to contribute to the difference in stability between the two proteins. To identify the residues that are responsible for the instability of DDX3Y, we leveraged the high sequence identities between DDX3X and 3Y. Sequence alignment showed two lysine residues on DDX3Y, K250 and K546, which are replaced by arginine residues in the same positions of DDX3X (Figure 4C). We therefore substituted these two lysine residues on DDX3Y with arginine individually. Both the K250R and K546R mutants were expressed at much higher levels than wild-type DDX3Y when U87MG cells were transfected with equal amount of plasmid encoding each protein (Figure 4D), indicating that both residues contribute to the instability of DDX3Y. Combining these two mutations did not further stabilize DDX3Y (Figure 4E). In addition, treatment with the proteasome inhibitor MG132 protected wild-type DDX3Y but not the K250,546R mutant (Figure 4F). Together these results suggest that K250 and K546, which are unique to DDX3Y, target this protein to the proteasome-mediated degradation pathway in U87MG cells.